Intel/roberta-large-mnli-int8-static-inc

INT8 RoBERT large finetuned on MNLI

Post-training static quantization

This is an INT8 PyTorch model quantized with Intel® Neural Compressor.

The original fp32 model comes from the fine-tuned model roberta-large-mnli.

The calibration dataloader is the train dataloader. The default calibration sampling size 100 isn't divisible exactly by batch size 8, so the real sampling size is 104.

The linear modules roberta.encoder.layer.16.output.dense, roberta.encoder.layer.17.output.dense, roberta.encoder.layer.18.output.dense, fall back to fp32 for less than 1% relative accuracy loss.

Evaluation result

	INT8	FP32
Accuracy (eval-acc)	89.8624	90.5960
Model size (MB)	381M	1.4G

Load with Intel® Neural Compressor:

from optimum.intel import INCModelForSequenceClassification

model_id = "Intel/roberta-base-squad2-int8-static"
int8_model = INCModelForSequenceClassification.from_pretrained(model_id)